#leave out possible long anomalies X_train = X_train[:int(len(X_train)*0.97)] Y_train = Y_train[:int(len(Y_train)*0.97)] batches_X,batches_Y = chunks(X_train,Y_train, 30000) print('Building model...') model = Sequential() model.add(Embedding(max_features_X, embedding_size, mask_zero=True)) for l in range(nb_layers): model.add(LSTM(embedding_size, hidden_size, return_sequences=True)) model.add(TimeDistributedDense(hidden_size,max_features_Y)) model.add(Activation('time_distributed_softmax')) if os.path.exists(fdir+'/weights.hdf5'): model.load_weights(fdir+'/weights.hdf5') print (model.shape()) rmsprop=RMSprop(lr=0.0002, rho=0.99, epsilon=1e-8, clipnorm=5) model.compile(loss='categorical_crossentropy', optimizer='rmsprop') if (mode =='train'): #save all checkpoints checkpointer = ModelCheckpoint(filepath=fdir+"/weights.hdf5", verbose=1, save_best_only=False) history = LossHistory() sample = Sample() print("Training...") for e in range(nb_epoch): print("epoch %d" % e) #for X_batch,Y_batch in zip(batches_X,batches_Y): for i, batch in enumerate(batches_X):
#leave out possible long anomalies X_train = X_train[:int(len(X_train) * 0.97)] Y_train = Y_train[:int(len(Y_train) * 0.97)] batches_X, batches_Y = chunks(X_train, Y_train, 30000) print('Building model...') model = Sequential() model.add(Embedding(max_features_X, embedding_size, mask_zero=True)) for l in range(nb_layers): model.add(LSTM(embedding_size, hidden_size, return_sequences=True)) model.add(TimeDistributedDense(hidden_size, max_features_Y)) model.add(Activation('time_distributed_softmax')) if os.path.exists(fdir + '/weights.hdf5'): model.load_weights(fdir + '/weights.hdf5') print(model.shape()) rmsprop = RMSprop(lr=0.0002, rho=0.99, epsilon=1e-8, clipnorm=5) model.compile(loss='categorical_crossentropy', optimizer='rmsprop') if (mode == 'train'): #save all checkpoints checkpointer = ModelCheckpoint(filepath=fdir + "/weights.hdf5", verbose=1, save_best_only=False) history = LossHistory() sample = Sample() print("Training...") for e in range(nb_epoch): print("epoch %d" % e)